Download data and study materials from OSF
This experiment was fielded as part of a TESS telephone survey. Data and materials for all the studies included on this survey is available here.
Principal investigators:
Yue (Jen) Shang
Indiana University
Email: jenshang@indiana.edu
Homepage: http://www.indiana.edu/~spea/faculty/shang-jen.shtml
Francesca Gino
Harvard University
Email: fgino@hbs.edu
Homepage: http://francescagino.com/
Rachel Croson
University of Texas at Dallas
Email: crosonr@utdallas.edu
Homepage: http://cbees.utdallas.edu/~crosonr/
Sample size: 505
Field period: 11/22/2005-2/12/2006
On a daily basis, people rely on others’ advice to make judgments. In two studies, we examine the differential impacts of similarity between the source of the advice and the person making the judgment in two settings: judging others’ behavior and judging one’s own actions. We find that similarity interacts with the target of the judgment. In particular, information received from a different advisor is more heavily weighed than from a similar advisor in judging others’ actions, but information from a similar advisor is more heavily weighed than from a different advisor in judging one’s own.
(1) People will assign a higher weight to advice provided by similar others than to advice from different others (hypothesis 1: Main effect of similarity).
(2) When asked questions about their own behavior, people will weigh advice from others less than when asked questions about others’ behavior (hypothesis 2: Main effect of judgment type).
(3) When asked questions about their own behavior, people will weigh information from similar advisors more than dissimilar advisors. However, when asked questions about others’ behavior, people will weigh information from dissimilar advisors more than similar advisors (hypothesis 3: Interaction effect between similarity and judgment type).
The study is concerned with how the similarity between a participant (in the role of advice-recipient and judgment-maker) and an advisor impacts the judgments about one’s own behavior and others’ behavior. We use a 2 (similarity: similar vs. different advisors) x 2 (judgment type: judgments of one’s own behavior vs. judgments of others’ behavior) design. The first factor is between-subjects, so each participant receives advice from either a similar or a different advisor. The second factor is within-subjects, so each participant makes judgments both about others’ behavior and about their own future/hypothetical behavior. We use this design to test our three hypotheses above.
Our dependent variable of interest (weight of advice, WOA) was derived from the differences between the first and second judgment formed. Specifically, the WOA is a measure of how much the advice is incorporated into the new judgment. In our setting: WOA = ((|Phase II judgment - Phase I judgment|)/(|Advice - Phase I Judgment|)).
In total, there were 839 valid data points for the WOA measure. These values were used in the analyses reported below. First, we compute the mean for each participant’s WOA values in each condition, and then we compare the distribution of those values across conditions. To assess the impact of the experimental manipulations, we compare participants’ WOA values across conditions. In particular, the values for WOA were subjected to an analysis of variance in which similarity (similar vs. different advice source) served as a between-subjects factor, and judgment type (own or others’ behavior) served as within-subject factor.
Results reveal a significant main effect of judgment type, F (1,130) = 128.55, p < .0001, η2 = .50, supporting hypothesis 2: Advice is weighed more heavily when the participant is judging the actions of others (M = 0.56; SD = 0.28) than when they are judging their own future/hypothetical actions (M = 0.21; SD = 0.26). However, in contrast to Hypothesis 1’s prediction, we find no significant main effect for similarity (p = .80): Information from similar advisors is weighed the same as information from different advisors.
Finally, the results reveal a significant interaction effect between judgment type and similarity, F (1,130) = 3.97, p < .05, η2 = .03, thus supporting hypothesis 3. In particular, information received from a different advisor is more impactful (M = 0.58; SD = 0.29) than from a similar advisor in judging others’ behavior (M = 0.53; SD = 0.26), but information from a similar advisor is more impactful (M = 0.24; SD = 0.27) than from a different advisor in judging one’s own behavior (M = 0.18; SD = 0.25). Figure 1 presents the mean values of WOA graphically.
The results of our study are consistent with two of our three hypotheses. First, we do not find the hypothesized main effect of similarity of advisor on judgments. Second, we find that people use advice more heavily when making judgments about others’ behavior than when making judgments about their own future or hypothetical behavior. Third, we find the predicted interaction effect between similarity and judgment type: information received from a different advisor is more effective than from a similar advisor in making judgments about others’ behavior, but information from a similar advisor is more effective than from a different advisor in making judgments about one’s own behavior. We suspect that the lack of main effect of similarity is caused by the nature of this interaction effect.
The experiment was divided into two phases, each including two blocks of questions (four judgments of others’ behavior and four judgments of one’s own). In Phase I, participants were asked to form judgments on their own. In Phase II, another party was described as an expert, either similar or different from the participant, and they advised the participant on the appropriate judgment. Participants then formed a second judgment for the same 8 questions.
More specifically, at the beginning of Phase II, participants were told “I will now ask you to answer the same series of questions. Yet, this time you will be given an answer from a randomly selected expert. Her name is Mary [His name is Tom].” Then, when asked for a judgment, participants were told what the advisor suggested. For example, for question 1 the interviewer said: “Among people who donate money to NPR or a local public radio station, what do you think is the average contribution per month? Mary [Tom] states the monthly contribution is 10 dollars.”
The similarity manipulation influences the description of the other party in Phase II, based on the profile of the participant. Participants randomly assigned to the similar condition were given a description of Mary [Tom] with features similar to their own profile, including gender, geographical region, education, political affiliation, and age. Participants randomly assigned to the different condition were given a description of the advisor with features different than their own on the same dimensions (see Appendix A for more detail). The values used as advice in Phase II were randomly selected by the experimenter among answers of 50 people who had previously completed Phase I. They were held constant across participants and across the two conditions (similar vs. different advisor).
Gino, F., J. Shang, and R. Croson. 2009. "The impact of information from similar or different advisors on judgment." Organizational Behavior and Human Decision Processes 108:287-302.